Chimeric Filoviruses for Identification and Characterization of Monoclonal Antibodies

ABSTRACT

Recent experiments suggest that some glycoprotein (GP)-specific monoclonal antibodies (MAbs) can protect experimental animals against the filovirus Ebola virus (EBOV). There is a need for isolation of MAbs capable of neutralizing multiple filoviruses. Antibody neutralization assays for filoviruses frequently use surrogate systems such as the rhabdovirus vesicular stomatitis Indiana virus (VSV), lentiviruses or gammaretroviruses with their envelope proteins replaced with EBOV GP or pseudotyped with EBOV GP. It is optimal for both screening and in-depth characterization of newly identified neutralizing MAbs to generate recombinant filoviruses that express a reporter fluorescent protein in order to more easily monitor and quantify the infection. Our study showed that unlike neutralization-sensitive chimeric VSV, authentic filoviruses are highly resistant to neutralization by MAbs. We used reverse genetics techniques to replace EBOV GP with its counterpart from the heterologous filoviruses Bundibugyo virus (BDBV), Sudan virus, and even Marburg virus and Lloviu virus, which belong to the heterologous genera in the filovirus family. This work resulted in generation of multiple chimeric filoviruses, demonstrating the ability of filoviruses to tolerate swapping of the envelope protein. The sensitivity of chimeric filoviruses to neutralizing MAbs was similar to that of authentic biologically derived filoviruses with the same GP. Moreover, disabling the expression of the secreted GP (sGP) resulted in an increased susceptibility of an engineered virus to the BDBV52 MAb isolated from a BDBV survivor, suggesting a role for sGP in evasion of antibody neutralization in the context of a human filovirus infection.

IMPORTANCE The study demonstrated that chimeric rhabdoviruses in which G protein is replaced with filovirus GP, widely used as surrogate targets for characterization of filovirus neutralizing antibodies, do not accurately predict the ability of antibodies to neutralize authentic filoviruses, which appeared to be resistant to neutralization. However, a recombinant EBOV expressing a fluorescent protein tolerated swapping of GP with counterparts from heterologous filoviruses, allowing high-throughput screening of B cell lines to isolate MAbs of any filovirus specificity. Human MAb BDBV52, which was isolated from a survivor of BDBV infection, was capable of partially neutralizing a chimeric EBOV carrying BDBV GP in which expression of sGP was disabled. In contrast, the parental virus expressing sGP was resistant to the MAb. Thus, the ability of filoviruses to tolerate swapping of GP can be used for identification of neutralizing MAbs specific to any filovirus and for the characterization of MAb specificity and mechanism of action.

INTRODUCTION

The family Filoviridae is composed of the genus Ebolavirus, which includes Ebola (EBOV), Sudan (SUDV), Taï Forest (TAFV), Reston (RESTV), and Bundibugyo (BDBV) viruses, the genus Marburgvirus, which includes Marburg (MARV) and Ravn (RAVV) viruses, and the putative genus Cuevavirus, which includes Lloviu virus (LLOV) (1). All of these viruses, with the exception of TAFV, RESTV, and LLOV, are known to cause disease outbreaks in humans with high case fatality (2, 3). The recent outbreak of EBOV disease in Western Africa (4) demonstrated that filoviruses can cause large epidemics. In addition, identification of the “new” filoviruses BDBV and LLOV was reported as recently as in 2007 or 2011, respectively (5, 6), suggesting the possibility of the emergence or identification of previously unknown filoviruses.

For decades, no treatment demonstrated protective efficacy against filoviruses in the nonhuman primate model, which is considered the best model of filovirus disease predictive for a similar effect in humans. However, recently developed treatments based on polyclonal antibodies (7) or monoclonal antibodies (MAbs) (8–10) have shown impressive levels of efficacy in nonhuman primates. The development of MAb-based treatments and understanding of mechanisms of antibody neutralization of RNA viruses can be greatly facilitated by the development of reverse genetics systems, which allow recovery of recombinant viruses from DNA copies of their genomes or antigenomes. The advantages of such systems for work with polyclonal or monoclonal antibodies include the possibility of introduction of mutations in genes encoding major protective antigens, such as the glycoprotein (GP), which is the sole envelope protein of filoviruses. GP is expressed as a precursor protein that is cleaved posttranslationally to GP1 and GP2 subunits, and the mature integral membrane protein is present on the surface of viral particles as two disulfide-linked subunits (11, 12). The GP gene of ebolaviruses encodes two proteins: the full-length GP, which is a part of the viral particles and a type I transmembrane protein, and the secreted GP (sGP). It also encodes a much less abundant small soluble GP (ssGP). The GP gene does not have a continuous open reading frame (ORF), and thus the expression of full-length GP and ssGP result from transcriptional editing. In contrast, sGP does not require transcriptional editing and has an identical N-terminal part of GP, but a unique C-terminal part (13–15). Unlike EBOV, MARV GP genes have a continuous ORF that encodes full-length GP but not sGP (16, 17). Another advantage of reverse genetics techniques is the possibility to engineer viruses expressing green fluorescent protein (GFP) or another reporter protein to visualize infection. However, development of reverse genetics systems is very labor and time-consuming and is not always successful. To date, such systems have been developed only for three filoviruses—EBOV, MARV, and RESTV (reviewed in reference 18)—and are not available for the other filoviruses causing a severe disease, including SUDV, BDBV, or RAVV. In the absence of reverse genetics systems and/or biosafety level 4 (BSL-4) facilities required for work with filoviruses, researchers use various surrogate systems for characterization of filovirus-specific MAbs or investigation of various steps of filovirus life cycle involving GP. These systems include chimeric vesicular stomatitis Indiana viruses (VSV) with the G protein replaced with filovirus GP (19–21) and pseudotyped gammaretroviruses (22, 23) and lentiviruses (24–26), which have their respective envelope proteins replaced with a filovirus GP provided in trans. It is generally assumed that these surrogate systems can be used to accurately characterize neutralizing properties of filovirus polyclonal or monoclonal antibodies.

In the present study, we used the EBOV reverse genetics system to show that the virus can tolerate replacement of GP with its counterpart not only from heterologous EBOV but remarkably also from a more distantly related MARV and cuevavirus. The generated chimeric viruses were used, in parallel, with chimeric VSVs, for characterization of filovirus-reactive MAbs isolated from human survivors of previous BDBV or MARV infection. We show here that filoviruses, including the chimeric filoviruses, are more resistant to neutralization by many MAbs, compared to their chimeric VSV counterparts. These data suggest that chimeric VSVs may not be optimal for accurate quantification of neutralizing activity of filovirus antibodies. We also demonstrate the development of a high-throughput system suitable for functional screening and analysis of large panels of filovirus MAbs. Finally, we used the chimeric viruses to demonstrate that BDBV sGP serves as a decoy for a BDBV GP-specific antibody isolated from a BDBV survivor.

MATERIALS AND METHODS

Construction of chimeric EBOV viruses and biological filovirus isolates used in the study.

To construct the full-length genome cDNA of EBOV with its GP swapped with that of BDBV, SUDV or MARV, we used the plasmid carrying the genomic RNA of wild-type EBOV (pEBOwt) and its modified version with the transcriptional cassette encoding enhanced GFP added between the NP and VP35 genes (pEBO-eGFP) (27), which were provided by Jonathan Towner and Stuart Nichol (Centers for Disease Control and Prevention [CDC]). First, the pEBOwtΔBamHI-SbfI, AscI-PspOMI plasmid subclone was generated from the pEBOwt construct by consecutive two-step removal of BamHI-SbfI and AscI-PspOMI fragments; after pEBOwt digestion with the each pair of restriction endonucleases, the residual vector part was treated with the Klenow fragment of DNA polymerase I and self-ligated. The resulting subclone was subjected to mutagenesis using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) for the introduction of NheI and XhoI restriction endonuclease sites flanking the GP ORF. Additionally, the similar pEBOwtΔBamHI-SbfI, AscI-PspOMI subclone was generated with a BamHI restriction endonuclease site instead of XhoI downstream of the EBOV GP ectodomain encoding region corresponding to the genomic sequence of EBOV Ebola virus/H.sapiens-tc/COD/1976/Yambuku-Mayinga, nucleotides 6039 to 7987 (GenBank accession number NC_002549.1). To clone the GP ORFs of MARV, BDBV, and SUDV, Vero-E6 cells were infected with Marburg virus/H.sapiens-tc/UGA/2007/Kitaka-200702854, Bundibugyo virus/H.sapiens-tc/UGA/200706291/Butalya-811249, or Sudan virus/H.sapiens-tc/UGA/2000/Gulu-200011676 and subjected to RNA isolation after cell lysis in TRIzol reagent (Life Technologies, Carlsbad, CA). Total RNA was reverse transcribed, and PCR fragments corresponding to the GP ORFs of three different filovirus species were generated and cloned into pUC19 vector by SphI and XmaI sites. In order to perform subsequent cloning steps toward generation of full-length clones, the following restriction endonuclease sites were knocked down by the introduction of silent mutations: BamHI and XhoI in MARV GP ORF, ApaI and NheI in BDBV GP ORF, and ApaI, SacI, and XhoI in SUDV GP ORF. To disable the expression of sGP, the GP gene transcriptional editing site was modified from AAAAAAA to AAGAAGAA (antigenome DNA sense) by site-directed mutagenesis, resulting in a plasmid designated pUC19-BDBV-GPΔsGP. Next, the resulting pUC19 clones were used to PCR amplify BDBV GP ORF with the following primers: direct, ACAGTAGCTAGCAACACAATGGTTACATCAGGAATTCTACAATTGCCC; and reverse, ACAGTACTCGAGAAAAATTAGAGTAGAAATTTGCAAATACACAGCAGTGC. SUDV GP was amplified with the following primers: direct, ACAGTAGCTAGCAACACAATGGGGGGTCTTAGCCTACT; and reverse, ACAGTACTCGAGAAAAATCAGCAAAGCAGCTTGCAAAC. MARV GP was amplified with the following primers: direct, TCTAGCAGGCTAGCAACACAATGAGGACTACATGCTTCTT; and reverse, TCTAGCAGCTCGAGAAAAACTATCCAATATATTTAGTAAAGATACGACAA. Finally, the MARV GP ectodomain was amplified with the following primers: direct, TCTAGCAGGCTAGCAACACAATGAGGACTACATGCTTCTT; and reverse, AGTCACGTGGATCCAGTCGGATGTCCACCATTTACCACC. In these sequences, the NheI, XhoI, or BamHI restriction endonuclease sites are underlined, and the start of the GP ORF direct sequence and the end of the GP ORF or GP ectodomain complementary sequences are italicized). The PCR products were used to replace the EBOV GP ORF in pEBOwtΔBamHI-SbfI,AscI-PspOMI subclone with the ORF for the GP of BDBV, SUDV, or MARV by NheI and XhoI sites or for the replacement of EBOV GP ectodomain encoding region with that of MARV by NheI and BamHI sites, respectively. To generate the final full-length constructs, the ApaI-SacI fragments of the generated subclones were transferred to pEBO-eGFP plasmid for substitution of the existing EBOV GP ORF with an ORF encoding the GP of BDBV, SUDV, or MARV or to replace the EBOV GP ectodomain with a MARV GP ectodomain. For the construction of EBOV full-length clone with its GP replaced by that of LLOV, we first changed editing site in LLOV GP ORF from 8A to 7A (antigenome DNA sense) by mutagenizing pBsII SK (+)-LLOV GP plasmid (provided by Ayato Takada [28]) to make it identical to the original LLOV sequence (6). Also, two existing KpnI restriction endonuclease sites in LLOV GP ORF were knocked down by the introduction of silent mutations. The resulting LLOV GP ORF cDNA was PCR amplified using the following primers: direct, TCTAGCAGGCTAGCAACACAATGGTGCCCACCTACCCGTA; and reverse, TCTAGCAGCTCGAGAAAAATCATCGTGTTATTCTGCACA (the NheI or XhoI restriction endonuclease sites are underlined, and the start of the LLOV GP ORF direct sequence and the end of the LLOV GP ORF complementary sequence are italicized). It was then cloned into the pEBOwtΔBamHI-SbfI,AscI-PspOMI plasmid. The ApaI-KpnI fragment from the resulting subclone was transferred to the pEBO-eGFP full-length clone with one of its KpnI sites (in polymerase L ORF, nucleotides 14292 to 14297 in the EBOV genome) disabled by the introduction of a silent mutation for the substitution of the existing ORF of EBOV GP with an ORF encoding the GP of LLOV. The chimeric viruses Ebola virus/H.sapiens-rec/COD/1976/Yambuku-Mayinga-eGFP-BDBV_GP (referred to here as EBOV/BDBV-GP), its derivative Ebola virus/H.sapiens-rec/COD/1976/Yambuku-Mayinga-eGFP-BDBV_GPdelta_sGP (referred to here as EBOV/BDBV-GPΔsGP) that is deficient in the production of sGP, Ebola virus/H.sapiens-rec/COD/1976/Yambuku-Mayinga-eGFP-SUDV_GP (referred to here as EBOV/SUDV-GP), Ebola virus/H.sapiens-rec/COD/1976/Yambuku-Mayinga-eGFP-MARV_GP (referred to here as EBOV/MARV-GP), Ebola virus/H.sapiens-rec/COD/1976/Yambuku-Mayinga-eGFP-MARV_GPed (referred to here as EBOV/MARV-GPed), and Ebola virus/H.sapiens-rec/COD/1976/Yambuku-Mayinga-eGFP-LLOV_GP (referred to here as EBOV/LLOV-GP) were rescued as previously described (29) and propagated by two passages in Vero-E6 cell culture monolayers. The genomic RNA of all recovered viruses was sequenced using Illumina HiSeq 1000 sequencing system as previously described (30), and the 3′ and 5′ termini were sequenced by RNA circularization as previously described (31). The sequences were deposited in GenBank (accession numbers KU174137 to KU174142). Work with the filovirus full-length clones was performed in a laboratory approved by the National Institutes of Health (NIH) Recombinant DNA Advisory Committee. Generation of the chimeric viruses was approved by the University of Texas Medical Branch (UTMB) Institutional Biosafety Committee. Recovery of the recombinant filoviruses and all work with filoviruses were performed in the BSL-4 facility of the Galveston National Laboratory. The growth kinetics experiments on chimeric EBOV viruses were performed as previously described (29). BDBV and MARV were provided originally by the Special Pathogens Branch of the U.S. Centers for Disease Control and Prevention (CDC) and deposited at the World Reference Center of Emerging Viruses and Arboviruses housed at the Galveston National Laboratory, the UTMB at Galveston. BDBV isolate 200706291 Uganda was isolated originally from the serum of a patient during the first recorded outbreak caused by this virus (5) and passaged three times in Vero-E6 cells. MARV isolate 200702854 Uganda was isolated originally from a subject designated “patient A” during the outbreak in Uganda in 2007 (32, 33) and underwent four passages in Vero-E6 cells.

Immunostaining of chimeric EBOV plaques.

Vero-E6 cell culture monolayers were inoculated with dilutions of chimeric EBOV constructs, covered with 0.9% methylcellulose (Sigma-Aldrich, St. Louis, MO), and incubated at 37°C. At day 6 after infection, the overlay was removed, and the cells were fixed with formalin for 24 h, taken out of BSL-4, and blocked in 5% skim milk in phosphate-buffered saline (PBS) containing 0.1% Tween 20 (Sigma-Aldrich) for 1 h. Next, monolayers were incubated for 1 h at 37°C with the selected MAbs (1 μg/ml) in solution or with 1:1,000 dilution of rabbit polyclonal serum raised against EBOV isolate Mayinga or MARV isolate Musoke and then washed three times with blocking solution. Thereafter, the respective goat anti-human or goat anti-rabbit IgG antibodies conjugated with horseradish peroxidase (KPL, Gaithersburg, MD) were added at 1:1,000 dilution in blocking buffer, and monolayers were incubated for 1 h at 37°C and washed three times with PBS. Virus plaques were visualized by staining with the 4CN two-component peroxidase substrate system (KPL).

Chimeric VSV with filovirus GP proteins.

VSV/MARV-GP was constructed and recovered as previously described (34). VSV/BDBV-GP was provided by Thomas Geisbert (UTMB), and VSV/EBOV-GP (35) was provided by Heinz Feldmann (Rocky Mountain Laboratories, NIH, Hamilton, MT).

Plaque reduction assay.

For plaque-based neutralization assays, 150 PFU of filoviruses or chimeric VSVs were preincubated with various concentrations of MAbs in a 100-μl volume for 1 h at 37°C in triplicate and placed on monolayers of Vero-E6 cells in 24-well plates. After adsorption of the virus for 1 h at 37°C, the cells were overlaid with 1 ml of 0.9% methylcellulose in minimal essential medium (MEM) containing 10% fetal bovine serum (Quality Biologicals, Gaithersburg, MD) and 0.1% gentamicin sulfate (Mediatech, Manassas, VA), followed by incubation at 37°C. Chimeric VSV plaques were visualized by staining monolayers with 0.25% crystal violet solution in formalin on day 3, 4, or 5 after infection, and filovirus plaques were immunostained with rabbit anti-EBOV polyclonal serum on day 6 after infection as described above. Plaques were counted, and neutralization curves were plotted as percentages of reduction of plaque numbers compared to mock-neutralized virus.

High-throughput screening (HTS) neutralization assay.

A total of 400 PFU of recombinant EBOV expressing eGFP or EBOV/BDBV-GP were incubated with various concentrations of MAbs in black polystyrene 96-well plates with clear bottoms (Corning, NY) for 1 h at 37°C in MEM containing 10% fetal bovine serum and 0.1% gentamicin sulfate. Next, 4 × 10⁴ Vero-E6 cells in the same medium were added to the virus-antibody mixtures, followed by incubation at 37°C for 4 days. The fluorescence intensity of infected cells at a 488-nm wavelength was measured in triplicate using a 2104 EnVision multilabel reader (Perkin-Elmer, Waltham, MA). The signal readout was normalized to virus control aliquots with no MAb added and is presented as the percentage of neutralization.

Binding of antibodies to the recombinant GPs of various filovirus species.

BDBV, EBOV, SUDV, or MARV GPs or BDBV, EBOV, or SUDV sGPs were coated overnight onto 384-well plates (Thermo Scientific/Nunc, Waltham, MA) at 1 μg/ml in PBS. Plates were blocked with 25 μl of blocking solution/well for 1 h. Blocking solution consisted of 10 g of powdered milk (Bio-Rad, Hercules, CA), 10 ml of goat serum (Gibco, Waltham, MA), 100 ml of 10× PBS, and 0.5 ml of Tween 20 mixed to a 1-liter final volume with distilled water. Purified antibodies were applied to the plates at a concentration 10 μg/ml in blocking solution for 2 h. The presence of antibodies bound to both GP and sGP proteins was determined using goat anti-human IgG alkaline phosphatase conjugate (Meridian Life Science, Memphis, TN) and p-nitrophenol phosphate substrate tablets (Sigma-Aldrich), with the optical density read at 405 nm after 1 h.

RESULTS

Filoviruses are resistant to antibody neutralization compared to chimeric VSVs.

We used human antibodies to compare their ability to neutralize the filoviruses EBOV and MARV on the one hand or recombinant VSVs expressing the same filovirus GPs on the other hand (Fig. 1). We first tested a recombinant form of the EBOV-specific antibody KZ52, isolated from a phage display library constructed from antibody genes from a human survivor of natural EBOV infection (36), which is protective in the guinea pig model of EBOV infection (37). The antibody appeared to neutralize authentic EBOV less efficiently than VSV/EBOV-GP. Next, antibodies BDBV43 and BDBV52, isolated from a human survivor of BDBV infection (38) were tested. Again, authentic BDBV was less efficiently neutralized than VSV/BDBV-GP. Moreover, BDBV52 did not neutralize authentic BDBV at any concentration tested (up to 200 μg/ml) but effectively neutralized VSV/BDBV-GP. Analysis of neutralization by additional BDBV antibodies from the survivor suggested a similar pattern, i.e., less efficient neutralization of authentic BDBV compared to VSV/BDBV-GP (data not shown). Analysis of the antibody MR201 that we isolated previously from a MARV survivor (34) also demonstrated a great difference between the neutralization of authentic MARV and VSV/MARV-GP. For example, 50 μg/ml of MR201 neutralized only 7% of MARV but as much as 94% of VSV/MARV-GP. Taken together, these data suggest that authentic filoviruses generally are more resistant to neutralization by MAbs compared to chimeric VSVs, which can, as a result of these findings, produce greatly exaggerated data on MAb neutralization.

FIG 1.

Filoviruses are resistant to neutralization by MAbs. The percentage of neutralization of filoviruses EBOV, BDBV, or MARV or the corresponding pseudotyped VSVs by human MAbs isolated from survivors were determined. Asterisks indicate concentrations of MAbs that gave different (P < 0.05) percentages of neutralization for filoviruses versus pseudotyped VSVs.

Replacement of the EBOV GP with its counterpart from heterologous ebolaviruses results in viable chimeric viruses.

Infection of cells with the recombinant EBOV expressing eGFP from an added gene results in a bright eGFP fluorescence in infected cells starting at 36 to 48 h postinfection (27), which opens the possibility for use of this virus for development of an HTS of MAbs. As noted above, the genus Ebolavirus includes five virus species whose representatives EBOV, BDBV, and SUDV cause a severe disease in humans. We therefore attempted to generate recombinant replication-competent EBOV-derived viruses in which the GP protein was exchanged with that of BDBV or SUDV. We replaced the ORF of EBOV GP in the EBOV full-length clone with that of BDBV or SUDV and used the plasmids in virus recovery experiments, as previously described (27, 29) (Fig. 2A and D). Indeed, the experiments resulted in recovery of viable chimeric viruses expressing eGFP: EBOV/BDBV-GP and EBOV/SUDV-GP (Fig. 3A). Sequencing analysis revealed no adventitious mutations in the GP gene or elsewhere in the genome. We then compared the multistep growth kinetics of the recombinant viruses (Fig. 3B). Contrary to our expectations, EBOV/BDBV-GP replicated slightly faster than EBOV, and EBOV/SUDV-GP replicated much faster than EBOV (Fig. 3B). For example, on day 1 or 2, the respective titers of EBOV/SUDV-GP were greater than those of EBOV by 2.5 or 1.3 log₁₀, although the final viral titers of all viruses were comparable. These data suggest that EBOV can tolerate swapping of GP with its counterparts from heterologous ebolaviruses.

FIG 2.

Swapping of GPs or their ectodomains with their counterparts from heterologous filoviruses result in viable chimeric filoviruses. (A to C) Cloning strategy for generation of EBOV/BDBV-GP, EBOV/SUDV-GP, or EBOV/MARV-GP (A), EBOV/MARV-GPed (B), and EBOV/LLOV-GP (C) full-length clones. (D) Schematic representation of the recombinant eGFP-expressing filovirus constructs.

FIG 3.

Swapping of filovirus GP has only a minimal effect on the efficiency of viral replication. (A) Fluorescent plaques of recombinant wild-type and chimeric filoviruses on day 3 after inoculation of Vero-E6 cell culture monolayers. The micrographs were taken at ×10 magnification. (B) Growth kinetics of the EBOV chimeras in Vero-E6 cells inoculated at an MOI of 0.1 PFU/cell. The red curve corresponding to EBOV/BDBV-GP is completely hidden under the blue curve representing EBOV/MARV-GP since these two viruses have similar growth kinetics. Asterisks or pound symbols show differences (P < 0.05) in viral titers of the chimeric viruses compared to EBOV on days 1 to 6.

EBOV can tolerate swapping of the glycoprotein with that of MARV and LLOV.

The two members of genus Marburgvirus, MARV and RAVV, have only partial antigenic relatedness (39), further suggesting the need for development of a single filovirus platform capable of expressing the GPs from individual Marburg viruses for accurate characterization of antibodies. Filovirus GP is a type I transmembrane protein (16, 17), which interacts with VP40 (40); its transmembrane domain affects conformation of the protein (41) and is required for incorporation in viral particles (42). Since the MARV GP cytoplasmic tail (amino acids RIFTKVIG) has no similarity to its EBOV GP counterpart (amino acids KFVF), and the transmembrane domains exhibit only a limited similarity (16, 43), the compatibility between MARV GP with EBOV particles was difficult to predict. We therefore initially attempted to replace the EBOV GP ectodomain only (which represents the GP lacking the cytoplasmic tail and the highly conserved transmembrane domain) with that of MARV (Fig. 2B and D). As a result, a fully viable virus designated EBOV/MARV-GPed was recovered (Fig. 3A), which replicated at a rate similar to that of EBOV (Fig. 3B). Since EBOV easily tolerated swapping of the GP ectodomain, we next attempted to replace the whole GP with that of MARV (Fig. 2B). This approach also resulted in recovery of a viable virus (Fig. 3A) that replicated at a level similar to that of EBOV (Fig. 3B), suggesting that the interaction of the GP cytoplasmic tail with VP40 (the matrix protein) may be not highly specific. We also attempted to replace the whole GP with its counterpart from cuevavirus LLOV, whose RNA was identified in a dead bat in Spain (6), but the virus was never isolated. This experiment resulted in recovery of a viable virus designated EBOV/LLOV-GP (Fig. 2A and C), whose replication was slightly reduced compared to EBOV (Fig. 3B). Again, adventitious mutations were not detected in the GP or elsewhere in the genomes of EBOV/MARV-GP, EBOV/MARV-GPed, and EBOV/LLOV-GP. Thus, EBOV can tolerate swapping GP with its counterparts not only from heterologous ebolaviruses but also from more distantly related and distinct marburgvirus and cuevavirus.

The generated chimeric filoviruses produce plaques, which can be stained by MAbs specific to their glycoproteins.

We next tested the ability of MAbs isolated from filovirus survivors to bind to plaques formed by the constructed viruses. We tested the binding of MAbs BDBV43 or BDBV52 recently isolated from a survivor of BDBV infection (38) and MR78, MR235, or MR246 from a survivor of MARV infection (34) (Fig. 4A). The comparison was performed in parallel using rabbit hyperimmune serum against EBOV or MARV. We found that rabbit EBOV-specific immune serum was able to stain plaques not only for EBOV but also for all constructs generated, including EBOV/MARV-GP and EBOV/MARV-GPed. This observation can be explained by the contribution of the binding of antibodies in the immune serum that recognize internal EBOV proteins, such as NP and VP40. Interestingly, BDBV43 stained the three ebolavirus GP-based constructs with various intensities and also stained EBOV/MARV-GP and EBOV/MARV-GPed, though with a low intensity, whereas BDBV52 stained EBOV/BDBV-GP and weakly stained EBOV/MARV-GP and EBOV/MARV-GPed but not EBOV and EBOV/SUDV-GP. These data suggest the two BDBV MAbs interact with epitopes that are partially conserved across all or some members of the family Filoviridae. In contrast, MARV polyclonal antibodies stained plaques of EBOV/MARV-GP or EBOV/MARV-GPed but not the ebolavirus constructs. Again, similarly to polyclonal MARV antibodies, the three MARV MAbs stained only MARV, but not the ebolavirus GP-based constructs. The GP proteins of EBOV and MARV have considerable amino acid similarity (44), with several stretches of four or more identical amino acids located in the receptor-binding region of GP₁ and heptad repeats 1 and 2 in GP₂ (based on comparison of EBOV isolate Mayinga and MARV isolate Uganda, Fig. 4B). In contrast, the mucin-like domains of EBOV and MARV have almost no sequence similarity. Therefore, the two BDBV MAbs react with conserved epitopes, while the two MARV-specific MAbs and the polyclonal MARV antibodies are directed against more variable epitopes. These data demonstrate that chimeric filovirus constructs are useful in the characterization of antibodies specific for any filovirus species.

FIG 4.

Plaques of chimeric filoviruses can be immunostained by MAbs specific to ectodomains of their GP proteins. (A) Vero-E6 cell culture monolayers were inoculated with dilutions of the indicated viruses, covered with 0.9% methylcellulose, and incubated for 6 days. Viral plaques were immunostained as described in Materials and Methods. (B) At the top, the degree of difference between the amino acid sequences of BDBV, SUDV, or MARV GP versus EBOV GP was calculated as 1.00 − H, where H is the position homogeneity (53). At the bottom, parts of GP1 and GP2 are designated by various colors under the following abbreviations: SS, signal sequence; RBR, receptor-binding region; GC, glycan cap; MD, mucin-like domain; IFL, internal fusion loop; HR1, heptad repeat 1; HR2, heptad repeat 2; MPER, membrane-proximal external region; TM, transmembrane domain; and CT, cytoplasmic tail (adapted from reference 54). Note that the lengths of the proteins, as indicated in the plots, are greater than those of BDBV, SUDV, EBOV, and MARV GP due to the gaps introduced in the alignments.

Chimeric filoviruses are neutralized according to the GP ectodomain specificity of MAbs.

To determine the binding specificities of the MAbs used in the study, we tested a panel of MAbs for their ability to bind a panel of recombinant filovirus GP or sGP proteins (Fig. 5A). All MAbs tested, with the exception of BDBV43, bound exclusively to full-length GP of the targeted filoviruses, and all of the MAbs except BDBV43 and BDBV52 did not detectably bind to sGP. In contrast, BDBV43 bound to both the GP and the sGP of all three ebolaviruses: EBOV, BDBV, and SUDV, but not to MARV, whereas BDBV52 bound to both the GP and the sGP of BDBV only. We next tested the ability of selected MAbs to neutralize EBOV, BDBV, MARV, EBOV/MARV-GP, or EBOV/MARV-GPed (Fig. 5B). As expected, EBOV was neutralized effectively by MAb KZ52 and, to a lesser degree, by BDBV43 but not by BDBV41 or BDBV52. BDBV was neutralized by BDBV41 and BDBV43, but not by BDBV52, despite the fact that this MAb binds to BDBV GP. MARV and the chimeric viruses carrying MARV GP or its ectodomain were neutralized by MR78 but not the other MAbs. Thus, the GP ectodomain specificities of chimeric filoviruses determine the neutralization efficiencies of antibodies, which do not necessary correlate with the protein binding data. These data suggest that chimeric filoviruses are useful for highly specific antibody neutralization tests.

FIG 5.

Chimeric filoviruses are neutralized according to GP ectodomain specificity of antibodies. (A) Binding of filovirus-specific human MAbs to recombinant GP or sGP of the indicated filovirus species. (B) Percentages of neutralization of wild-type or chimeric filoviruses by various concentrations of human MAbs.

Use of the chimeric filoviruses for a high-throughput screening (HTS) assay of MAbs.

Neutralization tests of large panels of MAbs or multiple serum samples against multiple filovirus species by conventional plaque reduction assay, which must be performed in BSL-4 biocontainment, is laborious. We therefore tested the generated chimeric viruses as targets for an HTS neutralization assay in which eGFP fluorescence provided a measure of remaining infectivity and thus a measure of antibody neutralization. In preliminary experiments, we performed test neutralizations of various doses of EBOV/BDBV-GP ranging from 4 to 4,000 PFU with various concentrations of MAb BDBV41. After a 1-h incubation, virus-antibody aliquots were mixed with various amounts of Vero-E6 cells ranging from 2,500 to 40,000 and then added as a suspension to individual wells of 96-well plates. eGFP fluorescence was read on days 2, 3, 4, 5, and 6. As an example of our preliminary data, infection of 40,000 cells with 150 or 300 PFU resulted in a 2-fold difference in the signal on day 4 (Fig. 6A). In another example, residual infectivity in Vero-E6 cells was measured in a broad range of various concentrations of MAb BDBV41 mixed with EBOV/BDBV-GP in triple aliquots. Quantitation of eGFP signal on a fluorescence plate reader on day 4 postinfection demonstrated that the level of eGFP signal was inversely proportional to the amount of MAb added (Fig. 6B). Interestingly, based on visual examination of plates, this antibody inhibited or prevented spread of the viral infection, but did not completely eliminate initial infectious foci even at the greatest concentration tested, 200 μg/ml (Fig. 6C). Based on the optimization experiments, we found that inoculation of 40,000 cells with 400 PFU (multiplicity of infection [MOI] of 0.01) of EBOV/BDBV-GP gave the greatest possible dynamic range of the signal at various antibody concentrations (data not shown). Next, we used the optimized assay conditions to determine neutralization potency of BDBV41 in comparison to the classic plaque reduction assay with BDBV, VSV/BDBV-GP, or EBOV/BDBV-GP (Fig. 6D). As in the case of KZ52, MR201, and BDBV43 (Fig. 1), VSV/BDBV-GP was more easily neutralized by BDBV41 in classic plaque reduction assays. Use of EBOV/BDBV-GP instead of BDBV for the plaque reduction assay resulted in a similar neutralization curve, and use of EBOV/BDBV-GP in an HTS assay also resulted in a neutralization curve similar to that generated by plaque reduction assays with BDBV or EBOV/BDBV-GP. Thus, chimeric filoviruses expressing eGFP can be used as a substitute for their natural counterparts in plaque reduction assays, as well as being used in HTS assays to rapidly identify MAbs neutralizing individual filovirus species.

FIG 6.

Use of chimeric filoviruses for HTS of MAbs. (A) Comparison of the levels of fluorescence on day 4 after inoculation of cells with 150 or 300 PFU of EBOV/BDBV-GP. (B) Levels of eGFP fluorescence in triplicate Vero-E6 cell suspensions inoculated with 400 PFU of EBOV/BDBV-GP pretreated with various concentrations of BDBV41. (C) UV fluorescence microscopy of Vero-E6 cells inoculated with 400 PFU of EBOV/BDBV-GP pretreated with various concentrations of BDBV41; MAb concentrations and the levels of fluorescence for representative wells I to IV are indicated in panel B. (D) Comparison of the neutralizing activities of BDBV41 in plaque reduction assay with a biological isolate of BDBV, recombinant EBOV/BDBV-GP filovirus or pseudotyped VSV (VSV/BDBV-GP), and in HTS with EBOV/BDBV-GP chimera.

The sGP protein prevents virus neutralization by BDBV52.

We next tested the possibility that the generated chimeric viruses are useful for a qualitative characterization of MAbs. We previously demonstrated that the secreted G protein of respiratory syncytial virus, which, similarly to EBOV is a nonsegmented negative-strand virus, reduces the efficiency of virus neutralization by serving as a decoy for neutralizing antibodies (45). More recently, a similar immune evasion mechanism was demonstrated for EBOV sGP (26). However, these studies involved mouse polyclonal antibodies, and the BDBV sGP study involved a chimeric VSV expressing EBOV GP, and therefore the importance of this mechanism for pathogenesis of human disease remained unknown. The data presented here show that BDBV52 bound to both BDBV sGP and GP (Fig. 5A). We hypothesized that sGP serves as a decoy for BDBV52 and antibodies with similar epitope specificities, thereby preventing their ability to effectively neutralize BDBV. To test the hypothesis, we modified the GP gene of EBOV/BDBV-GP cDNA to disable the expression of sGP by mutating the transcription-editing site from UUUUUUU to UUCUUCUU (negative-sense RNA strand). As a result, the modified GP gene had the continuous open reading frame encoding GP only. The resulting virus designated EBOV/BDBV-GPΔsGP was recovered, and sequencing data confirmed the stability of the mutation and lack of any adventitious mutations in the genome (data not shown). sGP is not required for viral replication in cultured cells. To determine whether disabling the expression of sGP makes the virus more sensitive to neutralization by BDBV52, we compared the susceptibility of the two viruses to the antibody. We found that whereas EBOV/BDBV-GP was completely resistant to the antibody, EBOV/BDBV-GPΔsGP was partially neutralized at the highest antibody concentration tested, 200 μg/ml (Fig. 7A), suggesting that BDBV52 binds part of sGP shared with GP (Fig. 7B). These data demonstrate for the first time the ability of an ebolavirus to evade neutralization by a naturally occurring human MAb isolated from a survivor.

FIG 7.

The sGP protein reduces virus neutralization by MAb BDBV52. (A) Percent neutralization of EBOV/BDBV-GP or EBOV/BDBV-GPΔsGP by BDBV52 or BDBV41. Disabling of sGP expression makes the virus partially susceptible to BDBV52. In contrast, both viruses are equally susceptible to BDBV41, suggesting that the increased susceptibility of EBOV/BDBV-GPΔsGP to BDBV52 is not a result of the altered properties of viral particles. (B) Binding of BDBV52 to both GP and sGP. Parts of GP are designated in colors as for Fig. 4B. sGP shares with GP the N-terminal part, including SS, RBD, and most of GC, and also has the unique part (UP) at the C terminus (13, 14). The most likely location of BDBV52 epitope, identified by Flyak et al. (38), is indicated for both sGP and GP_1,2.

DISCUSSION

The recent devastating outbreak of EBOV in Western Africa (4) demonstrated the pressing need in development of means of treatments and prophylaxes of infections caused by filoviruses. To date, the greatest progress has been achieved, which requires detailed characterization of monoclonal or polyclonal antibodies. The data presented here suggest that surrogate systems such as chimeric VSV expressing filovirus GP may produce misleading results and therefore appear to be suboptimal for the reliable characterization of filovirus antibodies. The simultaneous circulation of multiple lineages of filoviruses in the same outbreak (46, 47), along with the emergence of “new” filoviruses (5, 6), and improved methods for isolation of MAbs from survivors (34, 38, 48), along with the utility of their use in therapeutics, strongly support the need for improved efficient means of rapid screening and characterization of large panels of MAbs. Although a previous study demonstrated the possibility of the generation of viable ebolaviruses with some genes replaced with counterparts from a heterologous ebolavirus (49), the present study demonstrates that EBOV can easily tolerate exchange of GP not only from ebolaviruses but also from more distantly related marburgvirus and cuevavirus. Each of the chimeric viruses easily tolerated and expressed eGFP from an added gene. Moreover, we show that chimeric filoviruses are useful for an HTS screening process to identify neutralizing MAbs. In addition, we show that binding to GP does not necessarily predict the ability of a MAb to neutralize a filovirus, the effect most likely related to a different conformation of GP in a free form and in a viral particle. Taken together, these results illustrate the practicality of a quick exchange of GP in EBOV with its counterpart from any circulating filovirus and the use of the resulting chimeric filoviruses for a rapid analysis of large panels of monoclonal or polyclonal antibodies.

The present study shows that filoviruses are more resistant to neutralization by MAbs than chimeric VSV with filovirus GP widely used for quantitative analysis of filovirus-specific MAbs. Although the exact reason for this phenomenon requires additional studies, we hypothesize that it may be related to a much greater length of filovirus particles, 1,028 nm for EBOV or 876 nm for MARV (50), compared to 175 nm for VSV (51). The greater length of filovirus particles suggests a greater number of GP trimers per filovirus particle, which require a greater number of bound MAbs to abrogate infectivity, compared to a chimeric VSV. However, this model cannot explain why for some antibodies, such as KZ52 and BDBV43, the difference between neutralization of a filovirus and the corresponding chimeric VSV is moderate, while for others, such as BDBV52 and MR201, the difference is dramatic (Fig. 1A). Most likely, other factors, such as epitope specificity of MAbs, also affect the observed difference.

A recently published study demonstrated that sGP interferes with the antibody-mediated neutralization of EBOV GP-lentivirus pseudotype by antisera from sGP- or GP-immunized mice (26). The importance of this mechanism in the context of human filovirus infections remains unclear. In a separate study we isolated BDBV52, an antibody from a survivor, which binds both recombinant GP and sGP but does not neutralize EBOV/BDBV-GP (38). We hypothesized that this MAb can neutralize virus in the absence of sGP. To test the hypothesis, we generated a derivative of EBOV/BDBV-GP that does not express sGP: EBOV/BDBV-GPΔsGP. Indeed, this virus was partially neutralized by BDBV52. These data represent the first demonstration that ebolavirus survivors have MAbs that bind sGP, which do not neutralize the virus but can partially neutralize the virus when sGP expression is disabled. Thus, expression of sGP might help the virus to evade effective antibody neutralization. Of note, a recent study demonstrated that disabling of the transcriptional editing site of the GP gene reduces EBOV virulence, but the expression of sGP per se did not affect it (52). We are unaware of any comparison of sGP expression by the various ebolaviruses; however, the identical GP transcription editing sites and the high levels of similarity of the polymerase (L) genes of EBOV, BDBV, and SUDV (5, 41) suggest that the levels of expression of sGP by these viruses and by EBOV/BDBV-GP and EBOV/SUDV-GP are comparable.

In summary, these data suggest that (i) chimeric VSV expressing filovirus GP may not provide an accurate prediction of the neutralizing capacity of antibodies, (ii) EBOV can easily tolerate exchange of GP from other ebolaviruses or the heterologous marburgvirus or cuevavirus, (iii) swapping the GP genes and expression of a fluorescent protein from a recombinant filovirus allows for the rapid screening of large number of antibodies specific for virtually any filovirus, and (iv) sGP serves as a decoy, reducing the effectiveness of virus neutralization during filovirus human infections.